Wednesday, June 13, 2007

Before you do the war, do the math.

The ever-popular Cosmic Log provides a link to a University of Georgia press release announcing UGA assistant professor Patricia L. Sullivan's publication of a paper in which she describes a statistical method for predicting the outcome of a war. She claims an eighty percent success ratio in using her method to "predict", ex post, the outcomes of the 122 conflicts that have ocurred since World War Two, and which have involved one of the great powers against a weaker adversary. She says her method shows that the probability of U.S. success in the first Gulf War, i.e. of realizing its objective of driving Iraqi forces from Kuwait, was 93 percent. Similarly, the chance of the U.S. achieving its primary objective in the second Gulf War, the removal of Saddam Hussein's government from power, was about 70 percent. However, the likelihood of success at the longer term goal of maintaining a stable Iraqi government is now only 26 percent, and likely to take ten years to accomplish.

Sullivan also concludes that the Soviet Union had only a seven percent probability of prevailing in its war in Afghanistan. The press release doesn't state what her finding was with respect to the U.S. Afghan war, in which the long-term outcome still seems in doubt. If I had to hazard a guess, I would say that the probability of success with respect to overthrow of the Taliban government was high, given the existence of allies on the ground there. The likelihood of long-term stabilization may be less.

Of course, to apply Sullivan's method ex ante, you need accurate information concerning the various factors that affect the statistical determination. If your intelligence is faulty, your prediction of your chance of success may be off the mark. In other words: garbage in, garbage out